On the Coefficient of Determination: Simple But...

An example of possible misIeading role of the basic charaeteristics of the c1assical LB regression analysis is given. Another example using high breakdown point estimators demonstrates that in the case of contaminated data various estimators may give considerably different estimates. Consequently, a simple solution, routinely leading to the true model, need not exist. A proposal of one possibility how to cope with the situation is given and another is refered to.